Forum Discussion
Model Builder - Image classification - shifting from CPU to GPU
I'm trying to classify images based on 14 different categories. Currently I have about 40.000 images, but I'm planning to add more to try and get a better dataset.
Initially I have trained my dataset with my CPU and has pretty decent accuracy (90-98% in most cases), but the training and prediction speed is rather slow. Now I'm trying to use my GPU to speed up this task. The speed is amazing when it's training through the model builder, however the accuracy is terrible. I get about 8% accuracy and the expected category isn't even listed in some cases.
The only thing I changed is the CPU to GPU in the Model Builder GUI. Is there something I'm missing here? Do I require even more images to get the same results with my GPU vs my CPU?
I'm at a loss here so any help would be really appreciated!
2 Replies
- LuisQuintanilla
Microsoft
Hi FaithlessDbo,
Sorry to hear you ran into this issue. While it's possible to experience some variance in results, this seems like a large discrepancy. Can you please post your issue in the dotnet/machinelearning-modelbuilder repo so we can investigate.
https://github.com/dotnet/machinelearning-modelbuilder/issues/new/choose
ThanksThanks LuisQuintanilla I have made a post in your suggested repo. I hope it's possible to identify the issue.
